We describe the development and implementation of a new open configuration magnetic resonance imaging (MRI) system, with which neurosurgical procedures can be performed using image guidance. Our initial neurosurgical experience consists of 140 cases, including 63 stereotactic biopsies, 16 cyst drainages, 55 craniotomies, 3 thermal ablations, and 3 laminectomies. The surgical advantages derived from this new modality are presented.
The 0.5-T intraoperative MRI system (SIGNA SP, Boston, MA), developed by General Electric Medical Systems in collaboration with the Brigham and Women's Hospital, has a vertical gap within its magnet, providing the physical space for surgery. Images are viewed on monitors located within this gap and can also be acquired in conjunction with optical tracking of surgical instruments, establishing accurate intraoperative correlations between instrument position and anatomic structures.
A wide range of standard neurosurgical procedures can be performed using intraoperative MRI. The images obtained are clear and provide accurate and immediate information to use in the planning and assessment of the progress of the surgery.
Intraoperative MRI allows lesions to be precisely localized and targeted, and the progress of a procedure can be immediately evaluated. The constantly updated images help to eliminate errors that can arise during frame-based and frameless stereotactic surgery when anatomic structures alter their position because of shifting or displacement of brain parenchyma but are correlated with images obtained preoperatively. Intraoperative MRI is particularly helpful in determining tumor margins, optimizing surgical approaches, achieving complete resection of intracerebral lesions, and monitoring potential intraoperative complications.
The development of image-guided surgical methods during the past decade has provided a major advance in neurosurgery and may have major implications for other surgical fields as well. These methods allow greater accuracy in the localization of a lesion, the determination of its margins, and the optimization of a safe surgical approach. The use of image-based information represents a substantial improvement in the surgeon's armamentarium in the surgical treatment of tumors, vascular malformations, and other intracerebral lesions.
Present systems of image-guided neurosurgery include frame-based and frameless technologies (4,5,10,17,19,21,29,39,47,55). Both systems use images acquired preoperatively to create a three-dimensional space on which the surgical navigation is based. Frame-based systems, such as the Brown-Roberts-Wells (9,22), Cosman-Roberts-Wells(14), and Leksell systems(32-34) use externally applied frames to establish the fiducials for navigation. Frameless systems include those that use optical, electromagnetic, or ultrasound sensors and those using mechanical arms to track the position of surgical tools and instruments during surgical procedures (5,10,47,56,60).
All of these systems use images that are acquired preoperatively and therefore cannot provide the surgeon with information about dynamic changes that occur intraoperatively. The position of anatomic structures can change during surgery secondary to the egress of cerebrospinal fluid or the resection or biopsy of an intracranial lesion. Parenchymal shifts of greater than 5 mm during stereotactic biopsy have been documented(45) and can be even more pronounced in patients with hydrocephalus or substantial atrophy. The accuracy provided by both framed and frameless systems therefore decreases as a surgical procedure progresses. Furthermore, these techniques do not allow the detection of unexpected intraoperative events, such as hemorrhage.
During the past 7 years, physicians and scientists from the Brigham and Women's Hospital in Boston, MA, have collaborated with General Electric Medical Systems in the development of an open configuration magnetic resonance imaging (MRI) scanner that allows surgery to be performed with concurrent intraoperative image guidance. The initial research and development phase came to fruition in 1994 with the installation of a prototype unit at the Brigham and Women's Hospital (1,3,41). Since that time, there have been a number of initiatives exploring and evaluating the possible applications for this novel device.
We describe the concept of this new imaging system, its early development, and its implementation for intraoperative guidance. We present the initial neurosurgical experience at the Brigham and Women's Hospital with the intraoperative MRI scanner and provide some illustrative examples.
THE CONCEPT OF INTRAOPERATIVE MRI
The use of imaging modalities to provide guidance in the operating room is not a new concept. Image guidance has been applied for orthopedic procedures since the clinical introduction of the x-ray. Neurosurgeons have long used fluoroscopy during surgery, and more recently, intraoperative ultrasound has become a relatively routine procedure. Procedures performed with computed tomography (CT) were pioneered by Lunsford (37,38), and soon after the introduction of MRI, its potential for guiding biopsies, percutaneous thermal ablations, and other interventional procedures was recognized (8,12,13,20,24,31,36). A clear advantage of MRI over fluoroscopy and CT is that it does not expose the patient to ionizing radiation.
The use of MRI to provide intraoperative guidance for surgical procedures is, however, a relatively new concept (24) made feasible by the development of an appropriately configured, open access MRI system. This system incorporates several objectives. First, it represents an attempt to overcome the spatial restrictions that are imposed on neurosurgical approaches by the criteria of minimal invasiveness and that result in a limit to the neurosurgeon's extent of direct visualization. Intraoperative MRI provides the neurosurgeon with an opportunity to "see" beyond the exposed surfaces. Second, it provides a more sensitive method than direct visualization to distinguish diseased tissue from normal tissue. Finally, if direct visualization can be supplemented by real-time imaging, the changes in anatomy and tissue integrity that occur during surgical procedures can be monitored. The relatively high spatial and temporal resolution provided by MRI, in conjunction with the multiplanar and volumetric three-dimensional data that can be obtained, make interactive image plane definition possible. This can facilitate surgical localization and targeting of a lesion as well as improve intraoperative navigation.
MRI has demonstrated advantages over other imaging modalities in localizing tissue abnormalities and determining apparent tumor margins. It is therefore ideal for guiding various biopsies and tumor resections(15,42,48). The sensitivity of MRI in the detection of pathological changes is well established and, with the administration of contrast agent, tissue damage can be fully characterized. In addition, there is the potential to use MRI in the detection of subtle physiological, metabolic, or structural changes and to provide functional anatomic detail by evaluating parameters such as diffusion, perfusion, and/or flow.
Targeting and navigation
Magnetic resonance (MR) images acquired preoperatively aid in planning the optimal surgical approach to a lesion. Intraoperative MRI provides near real-time images, which enable the surgeon to correct or modify the preplanned trajectory of approach during the actual surgery. The use of light-emitting diode (LED)-based optical tracking of surgical instruments in combination with the manipulation of the MRI planes provides continuous interactive feedback between the surgical maneuvers during a procedure and the corresponding images (Fig. 1). This is the principal component of the image guidance in this intraoperative MRI system and is appropriate for the tracking and guidance of rigid instruments (i.e., needles, rigid endoscopes, and surgical instruments) (48,52). When using flexible instruments such as catheters, guidewires, and flexible endoscopes during intraoperative MRI, other nonoptical tracking methods based on MRI need to be used. Miniature coils and attached to these instruments can be detected by MRI when the instruments are placed deep within the operative field (35).
With both the optically based and the MR coil-based tracking methods, the acquisition and display of the corresponding images are versatile(16,35,52). The tips of the instruments being tracked can be superimposed and displayed on previously acquired images. Alternatively, real-time images can be taken to establish the position of a particular instrument. This position can be represented either in the reference frame of the magnet or in the moving reference frame of the instrument, allowing several options for interactive imaging and subsequent display.
The determination of the optimal trajectory for a particular target is one of the most important considerations in planning an intracranial procedure. With intraoperative MRI, images in three orthogonal planes are acquired to depict the anatomy. These can be obtained as the standard orthogonal planes or can be related to the axis of a proposed trajectory (Fig. 2). In the latter case, the position of the tracking instrument (e.g., needle holder) can be moved, allowing the surgeon to view and evaluate the various options for targeting. The three orthogonal image planes (axial, coronal, and sagittal) are related to the magnet, with the geometric center of the field of view representing the position of the instrument tip. The tip of the instrument being tracked can be used as an intersecting point between serial images and is thereby effectively used as a pointer(16). This pointer tip application is also used to assess the anatomic extent of tumors and identify tumor margins.
MR image sequences
The intraoperative MRI system can provide all conventional proton density, T1-, T2-, and T2*-weighted images with quality similar to that of a standard clinical 0.5-T system. A wide range of imaging sequences are available, but the most frequently used sequences for interactive scan plane selection are gradient echo (2-4 s/image), T1-weighted fast spin echo (FSE) (14 s/image), and limited field-of-view T1-weighted FSE (6-11 s/image). These sequences are obtained in 5-mm slices. For the frequent updating of an image volume, T1-weighted three-dimensional gradient echo sequences in 12 3-mm slices (59 s) and FSE T2-weighted sequences in 5-mm slices (16 s/image) are used. During laser hyperthermia, T1-weighted FSE sequences and phase-contrast imaging in conjunction with image subtraction are used to image local heat distribution and to depict the differences between images obtained before and during heat deposition.
HISTORY OF THE BRIGHAM AND WOMEN'S HOSPITAL INTRAOPERATIVE MRI DEVELOPMENT AND INSTALLATION
Phase I: development
The closed configuration of conventional MRI systems precludes direct access to the patient; therefore, fundamental changes in magnet and coil design, as well as display methods, were necessary to fully realize the concept of intraoperative MRI. The first interventional procedures in an open MRI system were performed in a low-field imager by Gronemeyer et al.(20). This system, like other low-field, open configuration diagnostic systems introduced later, provides some access to the patient through a horizontal gap in its magnet. Although some percutaneous interventions can be performed in this system, the access is limited and open surgeries requiring full access to the patient are not possible.
In 1990, our Chief of MRI (FAJ) had several conversations with representatives of General Electric Medical Systems to explore the possibility of an MRI device that could be used for real-time intraoperative guidance. Discussions with surgeons, including neurosurgeons (PMcLB, EA III), otorhinolaryngologists, and general surgeons, confirmed the potential importance of this concept. A team of engineers from General Electric Medical Systems and surgeons and radiologists from the Brigham and Women's Hospital was created for this purpose, and three neurosurgeons (PMcLB, EA III, PLG) provided input. After discussing several alternative designs, a "double doughnut" magnet system that would allow free access to the patient within the magnetic field was chosen for development. The system was first constructed as a mock up at the General Electric Corporate Research and Development Center in Schenectady, NY, and then further modifications were made to enhance the ergonomics for the subsequent design(48).
The prototype of this midfield intraoperative MRI system (General Electric 0.5-T Signa SP) was completed in 1994. Direct access to the patient was achieved by the construction of two vertically oriented superconducting magnets with coils in separate but communicating cryostats. This design results in a vertical gap between the coils, through which a patient can be fully accessed during image acquisition. The use of niobium tin, which has a maximum superconducting transition at higher temperatures than the more commonly used niobium titanium, allowed for sufficient cooling of the coils and thermal shield, with a two-stage cryocooler assembly (Gifford-McMahon, Kelcool UC130; Balzers, Hudson, NH), and eliminated the requirement for liquid helium baths. This elimination increased the area of access to the patient. Paired shielded gradient coils, separated to correlate with the access gap provided by the modified magnet, were developed concurrently. This new gradient coil design did not significantly reduce the gradient strength or linearity over the field of view. Another important feature of this system was the development of flexible transmit-receive coils, which can be incorporated into the sterile surgical draping and contoured over the surface of the patient's body in the area to be imaged, permitting access to the region of treatment beneath the coil (48).
The modified magnet provides a spherical imaging volume 30 cm in diameter and a 56-cm-wide area of patient access, which allows the surgeon and the first assistant to be positioned on either side of the patient. The physician can perform various percutaneous, interventional, endoscopic, or open surgical procedures while standing or sitting and can concomitantly view intraoperatively obtained MR images displayed on monitors, which are placed in the gap of the magnet. The position of the patient within the imager is flexible; the table can be inserted into the magnet along two orthogonal axes, allowing convenient access to different areas of anatomy. Alternatively, the table can be replaced by an adjustable chair, which permits the patient to be seated in several different positions. This flexibility provides several diagnostic imaging and procedural options.
After the initial stages of planning and development were completed, a full surgical suite, including a three-bed holding area for pre- and postoperative care, an instrument and equipment room, a scrub area, and a locker room, was built at the Brigham and Women's Hospital. Display monitors, surgical lighting, the anesthetic delivery system, and other requirements of sterile surgery needed to be integrated with the MRI system(Fig. 3).
The open intraoperative MRI system was installed in March 1994 at the Brigham and Women's Hospital and has been tested during a 2-year introductory period. Phantom studies using a variety of constructs were undertaken to establish the accuracy of the system. Biopsies of various anatomic lesions were then performed to assess the imaging and targeting features of the system, the effectiveness of tissue sampling, and the access to various parts of the body (52). Diagnostic applications included functional musculoskeletal studies and investigations of the loaded spine in various sitting positions (57).
Phase II: instrumentation
Many challenges needed to be met during the implementation of intraoperative MRI. These included the development of MRI-compatible equipment, instruments, and various tools along with the integration of the intraoperative display of images, the audio-visual communication between the team members, and the interactive manipulation of image data. The initial phase of MR-guided projects was hindered by the unavailability of MR-compatible surgical instruments. In many cases, extensive changes were required to adapt instruments and equipment to the unique electromagnetic environment (28). Presently, the lack of widely available MR-compatible instruments and equipment is still one of the most inhibiting factors in the development of interventional MRI. The parallel development of intraoperative and interventional MRI applications by other manufacturers may potentially motivate surgical instrument manufacturers to create more MRI-compatible surgical instruments.
This phase was one of experimentation, during which two problems were recognized and needed to be solved. Clearly, instruments that were ferromagnetic had to be assiduously avoided. However, supposedly nonferrous items could also be deflected within the magnetic field, and therefore, all potential instruments had to be tested. In addition, a number of metallic instruments that were not ferromagnetic created a substantial artifact when placed near a target within the imaging field of view and could not be used. Finally, electrically active equipment, such as the electrocautery, could interfere with image acquisition.
Small surgical instruments, such as curettes, scissors, and microforceps, were among the most problematic. Our intention was that the intraoperative MRI operating suite should have all of the components of a standard operating room, and through persistent, occasionally painstaking trials, the required surgical instruments were slowly assembled.
Several other key components required particular attention and adaptation. These included the headholder, the stereotactic navigating device, the bipolar coagulator, the drill, and the microscope. The headholder(Mayfield; Ohio Medical Instrument Co. Inc., Cincinnati, OH) was modeled on the three-point fixation system used in standard neurosurgical operating rooms. This unit, with a base that is fixed to the movable MRI patient table, needed to be designed to fit within the core of the magnet, accurately hold the patient in the desired position, and allow for adjustments as required.
The stereotactic navigating device was an essential keystone of the system and posed a greater problem. The initial testing of an optical tracking system was performed in a conventional system (FAJ, RK, and, from General Electric, William Lorensen). Two of our authors (EA III, RK) convinced General Electric to build a system directly into the intraoperative MRI that would allow an LED-based navigational system to be used with the center of the magnet as its three-dimensional space. Three linear charge-coupled device video cameras (Integrated Technologies, Boulder, CO), which register the movement of the LEDs within the imaging space, were fixed to the bore of the magnet. Several instruments with LEDs implanted were designed and used in conjunction with a flexible Bookwalter arm (Codman Inc., Burlington, MA) to serve as the tracking devices. Software analysis, using an interactive Sun 4/670 workstation (Sun Micro-systems, Mountain View, CA), allows the comparison of the location of these LEDs within the three-dimensional space. Used in conjunction with image acquisition, the surgeon achieves near real-time localization. The implementation of this was extremely successful and has provided precise navigation, with the tracking tip being accurate to within 1 mm (Fig. 1A).
An MR-compatible microscope was developed by Studor Medical Engineering Ag(Rhenfaal, Switzerland) and was tested intraoperatively at the Brigham and Women's Hospital in June 1996. The microscope is configured to fit between the magnets and over the shoulder of the surgeon. It has a mobile stand and is pneumatically driven, allowing fine control with a mouth clutch. A beam splitter for intraoperative video and observer capabilities is being developed. Imaging without distortion can be performed with the microscope in position.
An MR-compatible nonirrigating bipolar coagulator (Malis) was originally provided by Codman Inc. (Burlington, MA). Currently, a Codman gravity irrigating system is being used. The MR-compatible drill was obtained from Midas Rex (Fort Worth, TX). MRI-compatible anesthesia equipment, although expensive, did not pose a development problem, because the Ohmeda Company(Madison, WI) had previously developed an MR-compatible ventilator (Ohmeda Excel 21-0). Patient monitoring devices were obtained from Bruker(Maglite; ODAM, Wissenbourg, France). Sterile surgical draping materials were designed by Baxter (Deerfield, IL). These were adapted from standard surgical drapes and are affixed to the internal aspects of the magnet, where the surgeon is positioned. This provides a sterile operating environment equivalent to that of conventional operating rooms.
Phase III: clinical implementation
The intraoperative MRI suite is a hybrid, combining elements of an interventional radiology unit, an MRI facility, and and operating room. In this setting, the communication between team members (surgeons, radiologists, MR technologists, nurses, and engineers) is extremely important. A radiologist (FAJ, RBS) attended all of the cases and contributed to both the planning and the intraoperative imaging stages of the procedures. It is the radiologist's responsibility to select the optimal imaging strategy and direct the surgeon's approach to the lesion based on the images acquired. The radiologist interprets the images obtained throughout a procedure to accurately evaluate the information acquired and provide optimal guidance for the surgeon. The nursing staff, MR technologists, anesthesiologists, computer scientists, and engineers were also integral to the development and realization of the clinical use of intraoperative MRI.
The system was tested and validated for neurosurgical biopsies (TM) using various phantoms and experimental set-ups. The first stereotactic biopsy using this intraoperative MRI unit was undertaken (EA III, TM) in June 1995. Since that time, 63 biopsies of lesions located in the brain stem, deep white matter, thalamus, basal ganglia, cerebellum, and cortex have been performed using intraoperative MRI guidance (EA III, PMcLB, TM) (Fig. 2). There were no significant complications except for one episode of hemorrhage, which was immediately recognized intraoperatively through imaging. This was dealt with by turning a flap and evacuating the clot, and the patient had no clinical sequelae secondary to this complication. This event clearly demonstrated the value of intraoperative MRI in monitoring a procedure.
The implementation of intraoperative MRI for craniotomy was a challenge, principally because of the requirement for a high-speed drill and an operating microscope that could be used in the magnetic field. The first craniotomy for resection of a vascular lesion was performed (PS, TM) in June 1996 for the excision of a cavernous malformation. The first craniotomy for resection of a brain tumor was performed (PMcLB, TM) in a patient with a medial temporal ganglioglioma in August 1996.
Subsequently, 53 additional craniotomies have been performed using this intraoperative MRI unit to treat lesions that included intrinsic tumors, meningiomas, cavernous malformations, and arteriovenous malformations. The greatest use has been for low-grade gliomas. Initial imaging at the onset of the procedure allows localization of the lesion in relation to the positioning of the patient's head. It also provides for the optimal planning of the skin incision, craniotomy flap, and trajectory of access to the lesion. Verification of the intracranial location and extent of the lesion is accomplished with MR images obtained in conjunction with the navigational device. The lesion is removed, and serial imaging is acquired as necessary to guide the resection. The margins of resection can similarly be assessed. The intravenous administration of paramagnetic medium can help to determine the extent of resection (Figs. 4–6).
Intraoperative MRI provides a method for decompression and drainage of a cyst under direct visualization. This technique was first used (PMcLB, TM) in May 1996, and a total of 16 patients have been evaluated with arachnoid cysts or other cystic lesions. The administration of air or contrast medium directly into a cyst outlines its cavity and can be an important component for the treatment of this entity using intraoperative MRI. Information regarding septations or loculations within a cyst and whether it is in communication with the subarachnoid space can be elucidated through this technique.
The application of intraoperative MRI to the spine was dependent on the development of adequate intraoperative spinal rongeurs. In October 1996, the first spine procedure using MRI guidance was performed (EJW, TM) for the evacuation of an extra-axial cervical cyst. Cervical decompression for myelopathy was performed shortly thereafter. The ability of the surgeon to evaluate the extent of neural decompression in cases of tumors and spondylosis will undoubtedly increase the safety and effectiveness of these procedures (Fig. 7). Active efforts are also underway to combine endoscopic spinal surgery with intraoperative MRI to enhance the minimally invasive potential of these novel approaches. The ability to place and image a patient in the sitting position for this MRI procedure may additionally provide new innovations in areas of spinal research and dynamic spinal surgery.
Before the implementation of intraoperative MRI, laser hyperthermia had been tried as a method for ablation of an intracranial lesion (FAJ, PMcLB). MRI is unique as an imaging modality in its ability to visualize temperature changes dynamically and to therefore provide a mechanism through which thermal ablations can be monitored and controlled. Using intraoperative MRI in conjunction with laser hyperthermia allows the surgeon to view the deposition of energy within the brain parenchyma while proceeding with an ablation. Imaging thermal changes may increase the safety and effectiveness of focused interstitial laser therapy, radiofrequency ablation, and cryotherapy (8,23,25,26). The first patient underwent open laser hyperthermia treatment (PMcLB, FAJ) in September 1996, and a total of three patients have undergone laser ablation of parenchymal tumors in the intraoperative MRI suite to date(Fig. 8).
MRI guidance with single photon emission computed tomography (SPECT) image fusion
Information from other imaging modalities, such as dual isotope SPECT, can be coregistered with the MR anatomic data during MR resection to enhance functional localization beyond that offered by MRI alone. Other preoperative image data sets that may be similarly linked to the intraoperative MR images include data sets obtained using CT, positron emission tomography, functional MRI, dynamic MRI, MR spectroscopy, and transcranial magnetic stimulation motor or speech mapping. Functional imaging using thallium-201 (T1-201) dedicated high-resolution brain SPECT (CERASPECT brain imager; Digital Scintigraphics, Cambridge, MA) assists in the assessment of brain tumor recurrence versus radiation necrosis after aggressive adjuvant treatment with radiosurgery or brachytherapy (2,7,11,27,30,43,44,46,49,53,58,59).
At present, this unit can rapidly contour, superimpose, view, and manipulate data sets from multiple imaging modalities, including images obtained from intraoperative MRI using software that performs several discrete steps: contouring, fitting, reslicing, and display and manipulation of results. Intraoperative MR images are superimposed, and multiple image stacks can be interactively viewed using color maps and various overlay schemes. The method is accurate to within 2.5 mm in x andy dimensions and within one slice thickness in the z dimension.
This group has recently reported the correlation of survival and histopathology at the time of subsequent surgery (craniotomy after radiosurgery or brachytherapy), with the results of dual isotope SPECT for a large population of patients treated for glioblastomas multiforme(50,51). The combination of this functional imaging modality with the intraoperative images obtained using intraoperative MRI is an example of this powerful adjunct in surgical guidance(Fig. 9B).
ADVANTAGES OF INTRAOPERATIVE MRI
The inaccuracies that can occur when using framed stereotactic systems have been recognized. There is both a conceptual and actual difference between the mechanical accuracy of a device and the application accuracy of the same device. Inaccuracies that can arise during the application of a stereotactic frame include distortions of the frame because of the weight exerted by the head, movement of the head within the frame, error in setting the coordinates, and the effects of imaging protocols (18,39,40). Frameless systems do not have the encumberances of a framed system and have the advantage of providing some interactive guidance. Errors during the registration process, however, can be a source of inaccuracy in these systems. Movement of the fiducials during an operation after registration is completed is one of the most serious potential problems, because the prior registration then becomes invalid. Inaccuracies related to imaging must also be considered, because there can be significant distortion of the image peripherally where the fiducials are applied, rendering stereotactic calculations incorrect from the onset, in some instances. Inaccuracies as large as 8 mm have been recently documented(54).
The major problem with both the framed and frameless stereotactic techniques is that they rely on preoperatively obtained imaging and do not provide real-time updated information. This was well demonstrated with both framed and frameless stereotactic devices by Nauta(45). He showed that the limiting factors in accuracy were not mechanical or imaging errors but rather errors secondary to position changes of the brain during a procedure. With CT and MRI performed under ideal circumstances, stereotactic localization discrepancies greater than 5 mm were observed. Loss of cerebrospinal fluid, insinuation of air into the subdural space, changes in the anatomic position of a lesion with decompression or drainage, and surgically induced edema or hemorrhage are all contributing factors in the shifting of anatomic structures of the brain during surgery.
Systems using ultrasound for intraoperative guidance(6) can provide real-time feedback information; however, they have certain limitations that intraoperative MRI does not have. The view obtained is restricted by limits in the ability to place and move the ultrasound probe on the surgical field. Ultrasound resolution is not high enough to be used in the localization of lesions less than 5 mm or to clearly define and guide the total resection of tumor margins.
Operating under interactive MRI guidance offers neurosurgeons several advantages over previously used stereotactic image guidance systems. 1) With intraoperative MRI, there is no need for fiducials and registration or for the transformation from one frame of reference to another. 2) Intraoperative MRI offers superior capabilities in localizing a lesion in three-dimensional space and allows this localization to be updated in a dynamic fashion as changes in the fluid and tissue compartments of the brain occur. This has major implications in the ability to obtain accurate biopsies or resections of small and/or deep lesions, in defining the margins of resection, and in identifying the position of normal anatomic structures. 3) Intraoperative MRI allows not only precise preoperative plotting of the optimal trajectory of a surgical approach but also dynamic guidance and verification of the operative execution of the chosen approach. Because the approach can be carefully planned and then executed under MRI guidance, the surgical exposure to a particular lesion can be minimized at each step, from the initial skin incision, bone flap creation, and dural opening to the resection of normal tissue, if required, to access a lesion. The maximal preservation of normal tissue, made possible with this system, may contribute to decreased surgical morbidity. 4) Intraoperative MRI can help with the identification of normal structures, such as blood vessels, encountered in the surgical field and with the relationships (and any changes in these relationships as surgery proceeds) of these structures to a particular lesion. Similarly, surveillance of intraoperative complications, such as hemorrhage, is possible and can directly affect the patient's outcome. 5) Intraoperative MRI provides tissue characterization, which can be used to help the surgeon distinguish normal from abnormal tissue during resection. The cross-sectional information provided by MRI can be used to direct the surgeon to portions of a lesion that may not be apparent because of inaccessibility from the surgeon's position or view or because of a similarity in appearance to normal tissue. Special features unique to MRI can be used to enhance the information available during a procedure and to guide therapy. These include flow sensitivity, evaluation of perfusion, administration of gadolinium to detect a breakdown of the blood-brain barrier, or communication between fluid compartments. The ability of MRI to detect temperature changes can be exploited to monitor and control thermal ablations. Images acquired intraoperatively can also be combined and correlated with the results of other studies, such as SPECT or positron emission tomography and previously performed CT and functional MRI. The ultimate goal is to combine preoperative and intraoperative image data to provide the surgeon with the most comprehensive information possible to use in surgical decision-making.
Considering the remarkable potential of intraoperative MRI for neurosurgical applications, it should be emphasized that an optimal magnet design, high-resolution image quality, and the incorporation of navigational methods are all necessary to capitalize on the advantages of this revolutionary tool. The intraoperative MRI system we have developed and used has combined these features and allows the performance of open surgical procedures without the need of patient repositioning. Furthermore, by using advanced navigational tools and computer technology, it represents an integration of frameless stereotactic methods with real-time interactive imaging. This midfield imager, although not having the resolution capabilities of a 1.5-T diagnostic MR imager, provides sufficient spatial and temporal resolution and image quality to adequately assess anatomy and pathology, monitor a surgical procedure, and make image-based decisions. The intraoperative use of this unique system is not limited to biopsies or limited access procedures; if the required instrumentation is available, the entire range of neurosurgical procedures can be performed.
The development of the intraoperative MRI system may also initiate a reevaluation of current framed or frameless stereotactic devices. The serial intraoperative images obtained with our system have shown striking changes in the shape and position of brain structures and significant displacement of brain parenchyma. These direct observations should prompt heightened caution when using navigational systems that guide surgical maneuvers with preoperatively acquired image data.
The development of intraoperative MRI has been an important example of combining vision and creative thinking with interdisciplinary work. The result is a remarkable tool for neurosurgical applications. It allows surgical manipulation under direct visualization of the intracranial contents through both the eye of the surgeon and the volumetric images of the MRI system. This technology can be applied to cranial as well as spinal cases and can forseeably encompass application to the entire gamut of neurosurgical efforts. Our experience to date has shown that this device is easy and comfortable for the surgeon to use. Image acquisition, providing views in the plane of choice, takes no more than 14 to 60 seconds (depending on the imaging method) and therefore does not substantially increase the duration of a particular procedure. We think that the information received through intraoperative MRI will ultimately contribute to decreasing the duration of surgery and surgical morbidity.
Future possibilities include combining intraoperative MRI with other technologies, such as endoscopy and focused ultrasound, and the evaluation of brain functions intraoperatively. The routine use of this system in intracranial surgery may provide enough impetus to introduce its use in cranial base surgery, for which image guidance would be of essential importance and benefit. It is anticipated that this new development will result in additional innovations in surgical approaches and techniques.
We think that the development of intraoperative MRI marks a significant advance in neurosurgery, an advance that may revolutionize intraoperative visualization as fully as the operating microscope did. The combination of intraoperative image visualization and navigation is unparalleled and its applications to surgery are as broad as the field of surgery itself.
We express gratitude to the many people whose work and dedication made the clinical implementation of intraoperative MRI possible. In particular, we thank Amir Zamani, M.D., Liangee Hsu, M.D., and C. Winalski, M.D. (Brigham and Women's Hospital Department of Radiology), Marvin Fried, M.D., and Paul Morrison (Brigham and Women's Hospital Section of Otorhinolaryngology), George topulos, M.D. (Brigham and Women's Hospital Department of Anesthesia), Nobi Hata and Shin Nakajima (Surgical Planning Lab), Ray Kelley and Susan Koran (General Electric application specialists), Angela Ruddy-Kanan, Cathy Holley, and Carol Richards (nursing staff), Maureen Ainslee, Holley Isbister, and Cheryl Page (MR technologists). The Image-guided Therapy Program Director(FAJ) is supported by National Institutes of Health Grant PO1CA67165, National Science Foundation Grant BES9631710, DARPA Grant F41624-96-2-0001, General Electric Medical Systems, and Sun Microsystems.
This is the introduction of a new step in neurosurgical technique, combining intraoperative magnetic resonance imaging (MRI), instrumentation tracking ("neuronavigation"), and surgery in a specifically designed operating suite. This new technology was conceptualized by the Boston group at Brigham and Women's Hospital who present their concept, technical specifications, and first clinical experiences. The authors have achieved historical merit, having developed and improved this technology. The article concentrates on two topics: 1) instrumentation tracking for biopsies, and 2) intraoperative resection control by MRI.
1) The concept presented by the authors of combining a specifically designed open MRI magnet with neuronavigational techniques for the intraoperative tracking of an instrument is far beyond other current techniques of image-guided surgery for which preoperative images are used, implying the need for a reference (stereotactic frames or fiducials for frameless navigation) to register image space with physical space. Registration is not necessary with the open MRI scanner presented by the authors, because imaging and instrument tracking are performed within the same place. Because magnetic resonance (MR) images are acquired “online” during surgery, this concept is directly realizing the goal of “image-guided surgery” without having the potential disadvantages of the conventional neuronavigational techniques, such as brain shift or positional shift. Although intraoperative ultrasonography also has the capabilities of online imaging during surgery with instrument tracking, severe disadvantages of this method, specifically in discriminating tumor from normal brain tissue at the end of the resection procedure, remain.
The feature of MRI instrument tracking is predominantly used for stereotactic biopsies. MR image distortion as a major disadvantage of the conventional integration of MRI into stereotactic procedures may be avoided, because planning and needle targeting is performed within the same space. The advancing of the needle can be imaged online using the MRI double oblique slice positioning capability with instrument tracking using the neuronavigation device. This capability seems to offer major advantages concerning the clinical accuracy of biopsies and avoiding surgical morbidity(i.e., detection of hemorrhage).
2) The undisputed capabilities of MRI concerning soft tissue contrast and image resolution are used for intraoperative quality or resection control in low-grade or higher grade gliomas. Disruption of the blood-brain barrier at the resection margin remains important, introducing pitfalls concerning gadolinium enhancement. Image resolution limits caused by current low-field MRI need to be further evaluated. The indications for intraoperative resection control remains to be established, but based on our first experiences with intraoperative MRI control using a 0.2-T Magnetom Open, we are convinced that this novel technique will have its major advantages in glioma, pituitary, and epileptic surgery. Furthermore, the intraoperative use of functional brain imaging data will gain more and more significance.
The capabilities of the highly regarded Surgical Planning Lab augments the"hardware" concept presented by the authors concerning visualization and image fusion techniques that are necessary to take advantage of the intraoperative information provided by MRI for surgical strategies and intraoperative guidance. We are convinced that this new armamentarium opens a new world for operative neurosurgery.
Rudolf Fahlbusch Ralf Steinmeier Erlangen, Germany
MRI is generally accepted as the preoperative imaging modality of choice in modern neurosurgery. The value of early postoperative MRI has recently been well acknowledged (1). Intraoperatively, however, the surgeon still depends on clinical experience; in selected cases, the surgeon depends on ultrasound or, more recently, on computed tomography (CT)(2). All navigational devices, whether they are guided by mechanical arms, optical light-emitting diodes, or ultrasound, lack intraoperative correction of brain shift caused by spillage of cerebrospinal fluid, insertion of brain retractors, or resection of a lesion. Therefore, the use of intraoperative MRI, which can be directly compared to preoperative images, is a fascinating idea. The advantages of MRI, compared with CT and ultrasonography, are evident. The high tissue resolution allows the distinct determination of a lesion and its margins from the surrounding brain tissue. The ability of using nonreformatted three-dimensional images offers a superior way to plan surgical trajectories in open surgery as well as in minimally invasive procedures, such as biopsies and cyst drainages. The Boston group, consisting of neurosurgeons, radiologists, physicists, anesthesiologists, engineers, and technicians, made every effort to integrate an MR system in the operational environment. Although there exist also other concepts (3) of using low-field-strength magnets in the operating room, both systems have their advantages and limitations. Although the low-field systems have limited patient access, only intermittent scanning of the patients at certain stages of the procedure (e.g., resection control) is possible. Therefore, a sophisticated transportation system is necessary to safely reposition the patient before and after the scanning procedure. On the other hand, the surgeon has the comfort to perform the procedure with his usual instruments (e.g., the operating room microscope) and is not restricted by a gap between the two magnets. The concept described here makes the operating theatre a scanning room or vice versa. The high field magnets make additional safety procedures necessary, and the effect of long-term exposition to high magnetic field strengths has not yet been established. On the other hand, the spectrum of operating procedures, including spine procedures, is greater with a double-doughnut system. Nevertheless, this report represents a major step toward the ideal of having intraoperative on-line imaging control and will, in our opinion, move into many operating theatres. Both the systems and the concepts are not yet optimal, and a cost-effectiveness analysis needs to be conducted to determine the morbidity and survival time as well as quality of life for patients undergoing this procedure. The driving force of public opinion has to be managed carefully until the superiority of these systems is proven.
Volker M. Tronnier Heidelberg, Germany
Intraoperative MRI represents the next step in neurosurgery's constant search for a real-time navigation method with accurate intraoperative updating of the imaging data set. Black et al. describe their early experience with intraoperative MRI in a series of 91 cases, predominantly stereotactic biopsies and craniotomies.
The system as currently configured proved useful in the authors' hands for intraoperative estimation of the accuracy of biopsy, the extent of tumor resection, and even the detection of hemorrhage during one procedure. Although results are not provided for the entire series of patients, the illustrative cases aptly demonstrate the advantages of real-time intraoperative MRI in cases of tumors and spine dissections. It is surely most neurosurgeons' dream to obtain MRI confirmation of the extent of resection of a tumor or the degree of spinal decompression while still in the operating room.
As with any new technology, this initial report raises almost as many questions as it answers. How did intraoperative MRI affect the time course of the operation? Is it a time-saver because of smaller craniotomies and more accurate dissection or does the extensive involvement of the radiologist and the time involved in image acquisition and equipment setup lengthen procedures? Were there any consequences from field distortion inaccuracies affecting intraoperative guidance? How did dissection, hemorrhage, and tissue disruption affect gadolinium enhancement of tumor borders?
I think the answers to these and other questions will be supplied by this group and others working with intraoperative MRI in the near future. The rapid adaptation of any new technology reflects its true usefulness to the surgeon, its expense, and its ease of use. Frameless stereotactic systems have proliferated because they provide accurate navigation with an easy learning curve (1). Intraoperative MRI will add the advantages of real-time updating of the image set and immediate feedback regarding the success of the planned procedure. If these can be provided at a reasonable cost and with minimal change in the surgeon's routine, the technique will gain wide acceptance. At this early stage, intraoperative MRI needs to be refined by pioneers such as Black et al. so that all neurosurgeons will enjoy its benefits.
John G. Golfinos Patrick J. Kelly New York, New York
Black et al. pursued an evaluation of intraoperative MRI. They describe the pleasures and the pains of developing new technology and of trying to create an environment in which the union of surgical tools, imaging tools, and therapeutic tools can be accomplished. The authors raise many issues, which will likely be addressed by further studies at their institution and at other centers.
Eventual outcome studies will supplant the initial technological reports and allow us to more firmly grasp the overall importance of the technology. It is fundamentally clear that the union of advanced imaging tools and surgical techniques has produced a dramatic reduction in patient morbidity and has enhanced outcomes. Such technology has replaced previously invasive strategies with those that are safer and more effective. However, some questions remain.
What is the role of this particular tool within the spectrum of neurosurgery in comparison with existing technologies? Most image-guided procedures are performed most satisfactorily with preoperative image acquisition, either with frame-based or frameless techniques. I question the trade-off of patient positioning, availability of all necessary imaging and anesthesia tools, and working space for such a system.
Who should have the technology? Cost is an increasingly important issue, and before the technology is offered for dissemination, we should conduct studies that better define its role. The study presented by Black et al. provides a preliminarily persuasive rationale and an initial experience.
Is it cost-effective? In comparison with existing technologies that have been used in hundreds of thousands of cases, does this technology improve outcomes in such a way that the excessive costs of the technology are warranted?
What outcome studies are necessary to be able to show improvement in results? Is intraoperative MRI a technology looking for a substantive role?
Is the image resolution of the imaging technology satisfactory? In performing more than 3000 image-guided procedures at the University of Pittsburgh Medical Center, we have steadily progressed to more and more advanced imaging with higher resolution and greater detail: millimeter slice acquisition using 5/12 × 5/12 matrices, high-resolution monitors, multiplanar reformatted imaging, and image enhancement techniques with contrast and fat suppression sequences. All of these require the most high-resolution scanners. When we started with low-resolution MR scanners, we were distressed, thinking that the improvement over existing CT technology was poor. That is why we have switched the vast majority of our procedures to high-resolution MRI using 1.5-T magnets (and dramatically reduced the scan acquisition time). We now use a variety of surgical planning systems, both for radiosurgery and conventional stereotactic and functional neurosurgery.
Is the access to the patient really “free”? As depicted (demonstrated at meetings), it seems as though the thin neurosurgeon used to working with reduced instruments at his disposal will be happiest in this particular unit.
Is the instrumentation satisfactory for the procedure at hand or are we using suboptimal instrumentation to gain imaging compatibility? A balance must be achieved between the tools that are necessary to accomplish the surgical goal and the tools that must be used for satisfactory imaging.
Will this technology reduce risks? Black et al. note that 1 of their 50 patients undergoing biopsy techniques suffered a hemorrhage and advocate intraoperative MRI as a way to do without frame-based systems. Nonetheless, it was the introduction of frame-based systems that dramatically improved the results of biopsies, previously performed with freehand guidance. Probe stability and precision of the placement of the needle are the major attractive features of frame-based systems. Complications can be detected with a variety of alternative imaging techniques. We have treated more than 2000 cases with intraoperative CT, and our hemorrhage rate is 4 of 2000 cases. It will take extensive experience to show significant risk reduction for this technology.
Is brain shift important? The authors raise this as a significant issue, but there are very little supporting data. We have performed intraoperative CT for many years during tumor resections and have noted very little displacement or shift of the brain. Furthermore, the concept of maximal cytoreductive efforts for gliomas being enhanced by imaging, in our experience, proved not only impractical but impossible. After the entire contrast-enhancing portion of the tumor was removed and verified by intraoperative CT and the residual margins were biopsied, residual tumor cells were always identified.
These and other concerns will likely be addressed by future publications. The authors present a forthright analysis of the technology after pioneering its development. They and others will now be responsible for defining its true role.
L. Dade Lunsford Pittsburgh, Pennsylvania
In this report, Black et al. present a summary of their implementation of intraoperative MRI at the Brigham and Women's Hospital in Boston, MA. This unique project has been a highly productive collaboration between an Academic Health Center and an imaging device manufacturer, and it has yielded considerable results in both the academic and corporate business spheres. An overview of the project to date is appropriate; it has generated considerable interest in the neurosurgical community.
It is easy to appreciate why the capability of intraoperative MRI solves or sidesteps many problems for the current techniques of interactive image-guided neurosurgery. Intraoperative MRI is at such an early stage of development that much careful technical assessment and refinement is needed. Until the technology stabilizes, I suspect that many claims by its proponents will be difficult to confirm or disprove. For most neurosurgical practitioners, it will remain an intriguing idea undergoing rapid transition into clinical usefulness.
What is lacking in our current discussions about intraoperative MRI is an assessment of incremental clinical benefit versus capital expenditures plus marginal costs associated with these systems. Many academic centers engaged in research think that this device is interesting, but this curiosity is held in check by the formidable costs involved and the lack of obvious, measurable improvements in surgical outcomes resulting from its use.
In general, the increasing role for corporate sponsorship of neurosurgical research is a mixed blessing; it affords new opportunities for fruitful collaboration but opens us to many potential conflicts of interest. Therefore, it behooves us as investigators to be sensitive to our sources of bias during experimental design, performance, and reporting. Likewise, it requires us as readers to make an increasing commitment to critically assess the methodological details of publications in the neurosurgical literature. Meeting these challenges will ultimately enhance the opportunities for technological research and development while strengthening our role as independent evaluators of these efforts to improve clinical neurosurgery.
Robert J. Maciunas Nashville, Tennessee
From, Jacques Gautier D'Agoty, Essai d'Anatomie, en Tableaux Imprimes qui Representent au Naturel Tous les Muscles… de la Tête… d'après les Parties Disséquées & Préparées par Monsieur Duverney. Paris, 1745 (courtesy of Irwin J. Pincus, M.D., Beverly Hills, CA).